It has been suggested that computer simulations may be used for intelligence assessment. This study investigates what relationships exist between intelligence and computer-simulated tasks that mimic real-world problem-solving behavior, and discusses design requirements that simulations have to meet in order to be suitable for intelligence assessment. One hundred one participants took a test of inductive reasoning (BIS-K) and used the simulation MultiFlux [Kröner, S. (2001). Intelligenzdiagnostik per Computersimulation [Intelligence Assessment via computer simulation]. Münster: Waxmann.] designed to reduce the uncontrolled influence of prior knowledge, provide an evaluation-free exploration phase, and incorporate measures that are based on a theoretical model of simulation performance. Reliabilities of MultiFlux simulation performance scores were above .90, and the correlation of MultiFlux scores with BIS-K intelligence was, with r = .65 (adjusted r = .75), comparable to typical correlations among traditional intelligence tests. SEM analyses favored our theoretical performance model with three latent MultiFlux variables over a model with a single factor.